THURSDAY 10 SEPTEMBER 2026 latent·wire 82 PIECES ON FILE
← All subjects Subject · Models

Ling 2 0 Trillion Open Language Foundation

Covered by
arxiv.org

Timeline

  1. Ling 2.0 scales open MoE language models from 16B to 1T parameters

    Ling 2.0 is a claim that sparse MoE activation plus reasoning-focused training can push open language models to 1 trillion parameters without proportional compute, a scaling recipe other labs can test.

    1 source · thin-sourcing, excerpt-only

All sources (1)

  1. arXiv cs.LG2026-09-10

Related subjects